Deep Learning Based Multimodal Progression Modeling for Alzheimer's Disease. (3rd July 2021)

• Record Type: Journal Article
• Authors: Yang, Liuqing; Wang, Xifeng; Guo, Qi; Gladstein, Scott; Wooten, Dustin; Li, Tengfei; Robieson, Weining Z; Sun, Yan; Huang, Xin
• Is Part Of: Statistics in biopharmaceutical research. Volume 13:Number 3(2021)
• Page Start: 337
• Page End: 343
• Publication Date: 2021-07-03
• Subjects: Prognostic signature -- Structural magnetic resonance imaging -- Super Learner -- Transfer learning; Pharmacy -- Statistical methods -- Periodicals; Pharmaceutical biotechnology -- Statistical methods -- Periodicals; Biopharmaceutics -- Periodicals; Biometry -- Periodicals; Pharmacy -- Statistical methods; Periodicals; 615.190727
• DOI: 10.1080/19466315.2021.1884129
• Languages: English
• ISSNs: 1946-6315

Abstract

Abstract: The progression of Alzheimer's disease (AD) is a continuous process in cognitive and biomarker changes, with only a fraction of mild cognitive impairment (MCI) patients eventually advancing to AD. There is no definite biomarker signature to determine this eventual progress. The discovery and development of prognostic biomarker signatures for AD is essential to address the challenges of AD drug discovery and development. The deep learning (DL) technique is a recent breakthrough in data science that enables researchers to discover previously unknown features comprised by complicated patterns learnt from large datasets. It has outperformed many traditional machine learning methods in computer vision tasks. In this article, we evaluated the performance of DL algorithms in differentiating patients with diagnosis of AD, MCI, or no evidence of dementia using baseline MRI data of the brain, and integrated features extracted from the neural network with other baseline biomarkers to develop an AD prognostic signature. This signature can be used to understand the patient heterogeneity in a study cohort and provide enrichment strategies for AD clinical trial design.